Beruflich Dokumente
Kultur Dokumente
23. [ Introduction]
From Soft Architecture Machines
The design of structures for human activity is the basis of both architecture and human-computer interactionat least, if two senses of the word structures are considered. Conceptualizing a networked or software system as spatial makes connection between the two fields even more evident. This relationship is in fact borne out by the great contributions architects have made to new media. One of the most immediately obvious applications of virtual reality was in architecture, where the technology allowed for the first-person visualization of a planned physical building. The great cost of advanced graphics technology was justified by the much greater cost of construction and the need to get things right beforehand. This was one of the first advanced applications of the computer to design, and it lay outside the most typical fields in which early computing found application: accounting, the hard sciences, and engineering. Many new media insights offered by architects have come as innovative applications of architectural knowledge about space, or about design. The most influential architects working with computersincluding Michael Benedikt, Marcos Novak, William Mitchell, and Nicholas Negropontehave developed new principles and theories for the digital realm, both specifically with regard to the cyberspace concept and across new media in other ways. The idea that the user should be empowered by computers, rather than browbeaten into complying with a machine expert, is one particularly important idea that has been furthered by Negroponte. In 1967 Negroponte founded the Architecture Machine Group at MIT. It was in the context of working in this group that Soft Architecture Machines and his earlier The Architecture Machine were written. In this group, Negroponte and his collaborators developed methods of managing data spatially, rather than in the form of numeric or textual lists. The ideas laid out in the following two selections were, and remain, of great importance to the design of software. In Intentionalities, Negroponte describes three levels of awareness that a computer system should attempt, so as to be as responsive as possible. Even the most basic of these, having a model of the user, is absent from many ill-designed pieces of software today. (Consider how Microsoft Windows turns on the screen saver five minutes into showing a movie on DVD; even a primitive model of the user that knows about an activity such as viewing movie is absent.) In the following selection, Negroponte argues against the classical concept of the computer as an expert with special knowledge. Yet users carrying out tasks with computers today still often find themselves following along as software wizards direct user activity into one of a few pre-defined channels, in what is considered as a recent advance to improve productivity. Despite the power of Negropontes ideas, many that have proved themselves useful remain overlooked by software creators. With the support of MIT President Jerome Weisner, Negroponte continued and expanded his work by opening the MIT Media Lab in 1985, founding a unique institution. Sponsored by companies who get in-person access to students and faculty members, the Media Lab conducts research into future applications of technologies across many different academic disciplines, artistic media, and slices of life. Neither a corporate research lab nor a typical academic department, the Media Lab carries on the work of the Architecture Machine Groupespecially the approach of working with technologies more advanced than businesses will consider, and more unusual and yet more relevant to everyday life than the typical academic lab will. Negropontes influence has also been furthered through the magazine he co-founded, Wired, which sought to chronicle the digital revolution and promote it as a concept. His back-page essays from that magazine are collected in Being Digital. NM
As Negroponte indicates, similar ideas were used in an educational context by Seymour Papert, whose ideas became an important influence in Negropontes Media Lab, where Papert led a research group. See the selection from Paperts book (28).
353;
Materials on the CD illustrate some of the work done in the Architecture Machine Group; Richard Bolts paper (29) also represents a project of that group and deals with spatial data management, a concept later employed by Ben Shneiderman in his essay (33).
A Random Walk Through the 20th Century, by Glorianna Davenport and Michael Murtaugh is an interactive video documentary of Weisner and an example of some of the innovative work done at the Media Lab. That documentary, along with selections from the Media Lab videodisc Discursions, is included on the CD.
theNEWMEDIAREADER
Benedikt, Michael, ed. Cyberspace: First Steps. Cambridge: MIT Press, 1991. See particularly Michael Benedikt, Cyberspace: Some Proposals, 119224, and Marcos Novak, Liquid Architectures in Cyberspace, 225254. Brand, Stuart. The Media Lab: Inventing the Future at MIT. New York: Viking Penguin, 1987. Mitchell, William. City of Bits: Space, Place, and the Infobahn. Cambridge: MIT Press, 1995. Negroponte, Nicholas. The Architecture Machine. Cambridge: MIT Press, 1970. Negroponte, Nicholas. Being Digital. New York: Knopf, 1995. Original Publication Intentionalities60-63 and Computer-Aided Participatory Design, 102-123, Soft Architecture Machines. Cambridge: MIT Press, 1975.
Nicholas Negroponte
Intentionalities
I propose that a common oversight in the computer recognition and generation of visual material is the disregard for the intentions of the image. What I mean to say is more important than what I actually say. The intimacy of a dialogue can be in some sense measured by the ability of each person to recognize the intentions of the other. For example, in cases where people are not well acquainted and from different cultures, speaking to each other can become a profession (diplomacy) where it is very necessary to say exactly what is meant and to be well trained at understanding what is meant. With two good friends, codesigners, husband and wife, this is not true. A well-developed working relationship is in fact characterized by one partys leaving a great deal of information for the other party to infer and assuming it will be inferred correctly. As Oliver Selfridge puts it, an intimate, interactive conversation is, in some sense, the lack of it. Unfortunately, intentions can only be recognized in context, that evasive and omnipresent condition. But, in many cases, even the crudest definition of context (like now we are going to talk about structures in architecture) can help what Kaneff (1970) has titled The Picture Language Machine. If you are sketching a plan and I know
you are sketching a plan, even though some lines might replicate the schematic cat, I will do my best to assign to the lines a projective geometry or diagrammatic meaning associated with the built environment. However, if I know you are a lover of cats, there might be room (at some point) for equivocation, to the extent that I might have to ask, Do you mean . . .? There is nothing wrong with asking, but note that the need for asking is not necessarily a result of the level of detail, abstraction, or diagrammatic scribbling. The fact that most realistic rendering demands the same inference making and causes the same ambiguities is shown by trompe loeil painting and Ames experiments in the psychology of perception.
Figure 23.1. Examples of drawings made on the Architecture Machine as part of the so-called Cavanaugh experiment, designed to determine personalized drawing habits. Each figure is a computer display of every tenth point recorded by the data tablet. Figure 23.2. The Sylvania data tablet.
1975
To make inferences about a statement requires a knowledge of the world. To make an inference about the intention of a statement requires some knowledge of the person making it. For me to begin to make inferences about your intentionalities, except at the very crudest level (of contradictions, slips of the tongue, mispronunciations, etc.), requires that I know you (even as slightly as knowing that you are American). That is, I need a model of you. Following some work with Gordon Pask, we proposed in HUNCH An Experiment in Sketch Recognition (Negroponte, Groisser, and Taggart, 1972) that man-computer interactions should be supported by three levels of model. From the computers point of view, these include: (1) its model of you, (2) its model of your model of it, and (3) its model of your model of its model of you. The first level is a straightforward model of the user, ranging from his habits and mannerisms in sketching, for example, to his attitudes toward architecture. This model is continually exercised as a prediction device and supplier of missing information. Its validity is easily measured and tested in terms of the closeness of fit between the anticipated and the actual intention as manifest at some increment of time later (a millisecond, an hour, a year). Notice that in no sense can such a model be fail-safe; in fact, the very idea of fail-safeness itself is the wrong attitude toward the problem. In terms of implementation this model would be passive (and hence exhibit no inept behavior) at the beginning. After some period of time (with people this varies from personality to personality), this model is deployed to venture guesses and would inevitably make errors. Consider the process we go through in getting to know somebody. You will remember stages of attempting to make no predictions, times of wrong second-guessing, and later periods of knowing him or her. This is dramatically amplified if the other person is from another culture, illversed in your language. The next level of model is the computers model of your model of it. This is critical to inference making because one tends to leave implicit only those issues that one assumes the other party will understand (implicitly). This model grows out of a felicity of matches between the inferred information and the intended information. If, for example, the computer correctly assumes that you meant door within the wall, it can draw the added inference that you assumed it would. Note that this model can only grow out of correct matches.
theNEWMEDIAREADER
From aap noot mies huis by N. J. Habraken Amsterdam: Scheltema & Holkema, 1970. Translated from Dutch and abbreviated, from The Responsive House, edited by Edward Allen (Cambridge: MIT Press, 1974). Nowadays man lives in an unnatural relationship with his domicile. This artificiality becomes apparent when we know which types of natural relationships exist. There are six natural types of relationships. The seventh form of relationship brings into being non-homes. The first . . . is the simplest; the occupant builds his own house with his own hands. The second type of individual relationship is that in which the craftsman . . . offers his services. This relationship was very often responsible for housing in western history. The third type of individual relationship is that in which the architect acts as intermediary between occupant and craftsman. . . . There are very few who can afford this type of relationship. . . . The first collective type of relationship is that in which the community builds collectively the houses it needs, and does this without delegating the labor to craftsmen. The second collective type differs only by the delegation of some or all tasks to craftsmen. The third collective relationship is that in which the community and craftsmen do the actual building. The architect acts as the specialized intermediary. The seventh relationship is a nonrelationship. None of the previous types of relationship are found in mass production building. This seventh type is characterized by the fact that the occupants really take no part in it. They are unknown during the process of decision which leads to the production of dwellings. It is for this reason that . . . nothing reaches the architect from the group of the anonymous multitude of people. The architect is commissioned by another specialist who is no more the occupant than he is.
356 The concept further assumes that they can apply this understanding in concert with a competence to realize designs for the built environment. The results are an apparent (though not necessarily real) democracy in decision making, the consequence of which is ideally a responsiveness in architecture. This approach shortcircuits many of the traditional roles of the professional planner and architect regardless of whether he views himself as what Horst Rittel (1972) calls the doctor planner, the egalitarian planner, the needs planner, or the decisions planner. Consider two other examples of what can be viewed as the design of shelter: the design of automobiles and the design of clothes. In the case of the automobile most of us will agree that we personally do not know enough about combustion and mechanics to design our own cars. While exceptions like the Sunday mechanic and amateur car racer exist, most of us are satisfied with the existing selection of foreign and domestic cars, whether we view the automobile as a means to get us from here to there, as a status symbol, or as an extravagance. Therefore our participation in design is limited to supporting political lobbies to force Detroit to make cars safer. Clothes in some respect are at the other end of the spectrum inasmuch as I am confident that you and I can design and make our own clothes if we have to or want to. But clothes, unlike cars, require simple tools and involve materials that are generally easy to manipulate. At the same time, the low capital investment in materials and the high volume of the market allow for so many different kinds of clothing that anyone can find articles both that he likes and that are relatively unique within his circle of acquaintances. Note that our concept of fit is not demanding (most womens dresses come in only sixteen basic sizes). When we are fussy we can employ a tailor to make our clothes fit better though not necessarily to be better designed. Houses are somewhere between clothes and cars. They are not as expendable as shirts but are more manipulable than cars. There is a greater variety of kinds of houses than of cars, but any city offers less variety than the most meager haberdasher. The questions of this chapter focus on housing (which represents 85 percent of the built environment). The general thesis is that each individual can be his own architect. The participation is achieved in association with a very personal computing machine. Somewhat in contrast to Yona Friedman, I believe that a learning period with such a machine would be necessary, during which the machine would not make judgments and decisions but would ask telling and revealing questions and attempt to understand what you mean.
1975
357;
theNEWMEDIAREADER
358
Figure 23.12. The story of Mr. Smith: I had an idea about my house. I translated my idea into bricks. This is my house, the result of my translation. I made a mistake in translating, which I did not discover until I used my house.
Figure 23.14. The story of a neighborhood: Each of us had an idea about his house. We tried to explain our ideas to an architect, but there were so many of us that there was not enough time to explain our ideas sufficiently. The architect translated our ideas into an idea of his own. He liked his idea but we did not like it. And it is we who have to use these houses, not the architect! Figure 23.15. The story of another neighborhood: Each of us had his own idea about how to live. Our architect did not listen to us: he knew everything about the average man. The apartments he built were designed for the average man. But we are real people, not average at all. We are not comfortable living the way our architect likes to live.
Figure 23.16. A different kind of story: Each of us has his own idea about his house. Fortunately, there is a repertoire of all possible houses. Fortunately also, there are instructions about what to expect from each kind of house. Each of us can make his own choice, using the repertoire and the instructions.
Figure 23.13. The story of Mr. Wright: I had an idea about my house, and I explained it to the builder. The builder misunderstood me. The result is that my house has no door to the garden. Every time I want to use the garden, I have to get there through the window. My mistake was in not explaining more explicitly to the builder what I wanted him to build for me.
Figure 23.17. Each of us can thus plan the home of his choice, based on his own idea. In order to build our homes, we each need a lot, an access road, a water main, a power line, and so on. This is the infrastructure that supports each house. John wanted to build on lot 1. The others agreed . . . . . . After making sure that Johns choice of location did not hold disadvantages for them. Here the stories end.
1975
I remember one professor telling me that architecture is a form of social statement, that any building I ever designed ought to be the manifestation of profound symbolic comment. Isnt that both presumptuous and irresponsible, and, to say the least, paternalistic? While such attitudes may be applicable in a special context of building, I propose that they are generally inappropriate and a frequent cause of unresponsive architecture. The problem can be phrased in a simple question: Can an expert have expertise in goals and values, or is expertise per se limited to means? Father knows best for a long time. However, after a while he begins to lose credibility rapidly. Inconsistencies and unexplainable musts make the original institution of paternalism more and more suspect to a child; the doubt probably starts as early as age one or two. Nonetheless, for a long time the issue of Fathers rightness is less important than the comfort of knowing he is around. In this sense, it is interesting to question the role of the architect in terms of comfort and confidence; can it be embraced in a machine and thus avoid the potential orphanage of participation? Another question: If the architect as middleman is translating your needs in a built environment via transformation procedures seasoned by wisdom and his ability to pre-experience, what side effects and distortions take place in the process of this interpretation? How much of the deformity is positive in, for example, generating goals that you would never have thought of yourself? What do we lose when he goes away? Can a computer provide it? As a last question, consider the issue of risk. Can you seriously trust that someone who has no ultimate personal stake in the built artifact will do his utmost to achieve your personal and complex goals? An impelling motivation in most labors is in the consequence of doing a bad job. In contrast, the architect is released from all risk after his particular chunk of the built environment is built. The hazard to his reputation is slight, for he will be judged by colleagues and observers who do not have to live in what he has built and who will use extraneous criteria as the basis for criticism. In other words, the architect gets off scot-free, as innocent as the author of a bad novel.
359;
theNEWMEDIAREADER
360
buy a bigger one, then later, a smaller one. I can remember (but not reference) the statistic that the average American family moves every three years. The third observation, perhaps the most important, comes from my personal experiences of living on an Aegean island. It appears to be true that the local residents of an indigenous environment are unanimously dissatisfied with their architecture. Glass slabs are their metaphoric goals as much as, if not more than, the little white stucco house is mine. My electric typewriter has as much meaning as a Byzantine icon. Perhaps this can be explained in terms of communication technologies, by arguing that the local resident would be content, at a level to which we aspire, if he had not witnessed the electric toys of our times through magazines, television, and the passing rich tourist. However, a more deep-seated issue is the breadth of experience shared among these people. It is the case that they have in fact had very similar experiences among themselves and consequently carry nearly similar metaphors and share personal contexts. I am not saying that individuality has been squelched; I propose that the spectrum of experiences is small and may be accountable, in part, for this dramatic level of participation, so far not achieved in industrialized societies. It is quite clear that in faster-moving societies our personal experiences are phenomenally varied. This is why we have a harder (if not impossible) problem. This is why we need to consider a special type of architecture machine, one I will call a design amplifier.
Design Amplifiers
Before I begin I feel obliged to tell you that The Architecture Machine Group has worked very sporadically and without much success on this problem. The notion of a design amplifier is new and might provide an interim step between the present and the wizard machine, the surrogate human. I use the term amplifier advisedly; my purpose is not to replicate the human architect, as it may have been five years ago, but to make a surrogate you that can elaborate upon and contribute technical expertise to your design intentions. This allows us to consider and possibly see in the near future an option for computer-aided design that presumes informed machines, though not necessarily a machine intelligence. There is an inherent paradox here. A design amplifier will have no stake in the outcomes of joint ventures; hence it must act truly as an extension of the future user. Does this
1975
Figure 23.26. Mapping of preceding house plan Figure 23.27. A hexagonal resolution of the into planar graph. graph.
in turn mean that the machine intelligence necessary to support richness of dialogue will in fact be counterproductive to the participation because this same intelligence, like that of the human architect, would fall prey to the ills of translation, ascribing meanings of its own? In other words, does the intelligence required to communicate contradict the notion of informed amplification? I would draw your attention to the analogy of a good teacher who fosters an intellectual environment in which you discover for yourself in comparison to the one who drills facts and proclaims principles. As such, let us consider aspects of a design amplifier in terms of a somewhat dual existence: the benevolent educator and the thirsting student, all in one. There are two categories to consider: (1) What does the machine know? (2) How does the user deal with what it knows? These questions are particularly interesting because the most obvious paradigm is in fact the least rewarding. The most obvious method would be to construct a machine with a vast knowledge of architecture and to view the user as an
explorer of this knowledge through a window of his needs and the medium of some sophisticated man-machine interface. An example of this is found in most computeraided instruction systems where, for example, the machine knows arithmetic and the child manipulates the machine in a more or less prearranged exploration, witnessing yeses, nos, dos, and donts. A more exciting approach applicable to a design amplifier can be found in the recent work of Seymour Papert (1971a, b, c) and his colleagues. In brief, their theory is that computer-aided instruction should be treated as the amplification and enlightening of the processes of learning and thinking themselves, rather than merely presenting and drilling specific subject matter. To achieve this, the computer is treated, in some sense, as an automatic student by the child (see also Ackoff and Emery, 1972). In the Papert experiments, the six- or seven-year-old youngster has the opportunity to give a behavior to the computer via a simple but powerful programming language called LOGO. Whether
361;
theNEWMEDIAREADER
(Negroponte and Groisser, 1967a and b; 1970). However, it should be recognized that URBAN5 was the ultimate paternalist; it suffered from (among other things) being directed to serving the architect, not the resident.
Two Loops
Following Friedmans simple model of two loops, one with me and one with my neighborhood, lets examine some aspects of each in terms of automation. The reader should refer to the extensive works of Friedman, found in the bibliography, but should be cautioned about the particularly French notion of a banque de donnes or what he calls a repertoire. It is somewhat misleading taken at face value because it assumes a menu-picking activity rather than a design activity. The offerings of a menu of solutions obviously cannot exceed the combinatorial product of the parts (which may be enormous). Friedman, unlike many researchers in France, escapes this particular constraint by making his repertoire (banque de donnes) contain topologies that do not have a metric. It is the users adding of this metric that affords the limitless variety. The first loop is private. It must be self-sustaining in its powers to maintain the users attention, ask intelligent questions, and provide broad commentary. It must tread the thin line of distinction between making suggestions and being a bully, between criticizing and insulting, between navigating a search and directing it. Friedman (in personal conversation) makes a distinction by calling the computing organism a consultant, implying a knower at your beck and call, paid to help even if he may not agree with your personal premise. The connotations of consult are illuminating in the sense that the underlying skill is uncluttered by metaphorical distinctions, but it is also disturbing inasmuch as one must seriously question whether proficiency can be shared without dialogue requiring metaphor. The closest example of the first loop implemented in a computer program is ARCHIT (Rorick, 1971). The following is an excerpt from one conversation.
Computer: Shall we discuss the dining areas? User: Yes. Computer: Who does most of the cooking in your family?
362
1975
User: Carol. Computer: I would suggest that the dining area for your everyday meals be in the same room as the cooking area so that everyone can socialize while meals are being prepared and Carol wont be isolated in the kitchen. Dont you agree? User: Yes. Computer: How often do you want these occasions to be formal (that is, other than casually joining you for dinner) in times per year? User: 12. Computer: Keeping these answers in mind, do you feel that you need a separate dining area for more formal occasions? User: No. Computer: How many people do you want the everyday dining area to accommodate including company? User: 8.
363;
At this point the reader should be appalled. The computer program has browbeaten the user into not wanting a dining room and short-circuited the deeper patterns of living that may keep Carol in the kitchen quite happily away from everybody. The program has exhibited an illusion of intelligence and knowingness and, in this example, done all the talking! Unfortunately, I do not have a more positive example to offer (but am working on it). A blatant flaw in ARCHIT-like programs is the desire to rapidly pinpoint an architectural program via direct yes/no, one/two questions. Inference making and indirect procedures should be used, not for the purpose of making life difficult (for the computer), but for the purpose of soliciting more complex and revealing patterns of living. We must avoid initiating dialogue by asking questions because the questions perforce flavor the answer. The next section describes a simple experiment in inference making, one that avoids asking questions. In contrast to the inner loop, the outer loop is a great deal easier to conceive. Its purpose is to flag local perturbations when a desire of mine conflicts with an amenity of yours or of the group at large. A simple example would be a construction of mine blocking light or view from a portion of your house. Such functions assume that the machine is all-knowing about geometry, particular desires,
Plan Recognition
A typical exercise in computer-aided design is the generation of two- and three-dimensional layouts from a set of wellspecified constraints and criteria. The classical and most recent experiments can be found in Bernholtz (1969), Eastman (1972), T. Johnson et al. (1970), Liggett (1972), Mitchell (1972b), Mohr (1972), Quintrand (1971), Steadman (1971), Teague (1970), Weinzapfel (1973), and Yessios (1972). The underlying and common thread of all these works is the framework: input of problem specification and output of physical description. This section considers an experiment that seeks to do the reverse: input of a physical description (through recognition rather than specification) and output of problem specification. The goal is to recognize a structure of relationships and attributes in contrast to asking for a description.
theNEWMEDIAREADER
364
The initiation of the dialogue is achieved by mapping the physical plan into a relational structure (like the adjacent graph in figure 23.2) that does not have a metric (hence the initial unimportance of scale). The structure then is used to generate other solutions, assuming that the structure is underconstrained as a result of recognizing only a subset of the relations. It is much like only half-listening to a story, extracting an incomplete theme, and developing a new narrative (with similar structure). The other plans (that is, the machines story) reveal physical arrangements that have enough commonality for the user to make interested comments and for the machine to pose interesting questions. Interesting is defined here as leading to an increase in the users realizing and understanding architectural implications and an increase in the machines apprehension of the particular needs and patterns as manifest by what the user has now. The plan recognition program, SQUINT, employs the services of HUNCH. In particular, it exercises the feature of zooming in and out of the positional data, traveling within the spectrum of very low and very high resolutions. The preceding chapter illustrates the sort of range; the grain varies from 1,024 rasters per grain to a one-to-one correspondence. And, at any grain except the finest, the percentage of hits can be viewed as a gray tone. As happens with HUNCH, the noble intentions of SQUINT become reduced to very straightforward operations. Simple properties are recognized from the limiting boundaries of spaces and the penetrations of the boundaries. The first step is to look for the total number of bodies in the sketch. While there is usually one, this initial observation is necessary, if for nothing else than to save memory by compressing the positional data to exclude the white of paper that lies outside the sketched plan. The recognition of discrete bodies is achieved by a flooding process that creeps in from the sides of the paper, flowing around obstructing lines at a grain appropriate to ensure that it does not seep through doors and windows. Subsequent to flagging all flooded bits, the remainder are accounted for in a similar flooding technique, starting at any point. If all points are not accounted for by the first two floods, then there must be more than one body, and the procedure needs to be repeated until all points are tagged. It is the responsibility of later routines to decide whether the multiple elements in fact
1975
represent two autonomous disconnected sections of a house, for example, or whether in reality the additional figures are diagrammatic elements: north arrows, lettering, doodles, or coffee stains. Following the location of the silhouette(s) of the plan, rather similar procedures wander through internal subdivisions from one space to another, at one grain or another, a little bit like an expandable/shrinkable mouse meandering through a maze. Most sketching techniques will allow for internal spaces to be attained at the finest resolution. However, some sketching techniques include the demarcation of door radii and steps, which would impede passage of our mouse if the lines were considered boundaries (which they are not). These are the interesting cases; one must look for cues and develop evidence that, for example, such-and-such is probably a tread and not a chimney flue or this is probably a jamb and not a sill. Some of these situations are particularly difficult to deal with, where, for example, in one case the misinterpretation of a one-step level change resulted in guessing that the entire circulation of the house passed through the guest closet. This extreme example may appear to be a violent programming oversight. I must repeat, however, that there will always be conditions of such ambiquity that will require even the onlooking human to ask. I further insist there is nothing wrong with asking! Irrespective of whether the user has ascribed names to spaces, the program will give its own names in order to have an internal nomenclature of nodes and links. The labels can apply to traditional names (if you insist) like bathroom and bedroom; to orientations like north, windward, or vieworiented; or consist of schematic titles like space A, B2, or 732. The labeled nodes of the structure are linked with either categorical yes/nos or graded values of an attribute like access/circulatory, visual, acoustical. The subsequent mapping into an alternate floor plan has been done by Steve Handel and Huck Rorick. Roricks experiment appends the somewhat extraneous but interesting feature of adding heuristics that represent his view of what another architect might have done. In the specific case illustrated he has developed heuristics for overlaying a third dimension upon the plan following the vernacular of Frank Lloyd Wright, generating a variety of Wrightian roof forms. Though this is contradictory to the
full level of participation suggested by Friedman, it is fun to speculate that a representation of a deeper structure of my needs could be manipulated and displayed in the formal jargons of various famous architects, perhaps even Vitruvius or Viollet-le-Duc. We should not forget that the user of computer-aided participatory . . . is not an architect. Plan recognition might imply to some a more formal approach than is intended. The reader should be referred, if he is interested in the morphologies of floor plans, to the original works of Levin (1964), Whitehead and Eldars (1964), Casalaina and Rittel (1967), and the most recent work of Weinzapfel (1973). However, remember that these systems assume the driver to be an architect.
Bibliography Ackoff, Russell L, and Fred E. Emery. Our Purposeful Systems. Chicago: Aldine Atherton, 1972. Bernholtz, Allen. LOKAT. Laboratory for Computer Graphics and Spatial Analysis, William Warntz (editor). Cambridge, Mass.: Harvard University, IV.4, 1969 March. Casalaina, V., and H. Rittel. Morphologies of Floor Plans. Paper for the Conference on Computer-Aided Building Design, 1967. Cross, Nigel. Impact of Computers on the Architectural Design Process. The Architects Journal, 623, 1972 March 22. Davidoff, Paul. Advocacy and Pluralism in Planning. Journal of the American Institute of Planners, 31, 331338, 1965 November. Eastman, Charles M. Adaptive-Conditional Architecture. Design Participation, Nigel Cross (editor). London: Academy Editions, 5157, 1972. Friedman, Yona. Realisable Utopias. 1973. Friedman, Yona. Society=Environment. Brussels: C.E.A., 1972a. Friedman, Yona. Information Processes for Participatory Design. Design Participation, Nigel Cross (editor). London: Academy Editions, 4550, 1972b. Friedman, Yona. Flatwriter: Voice by Computer. Progressive Architecture, 52, 98101, 1971 March. Goodman, Robert. After the Planners. New York: Simon & Schuster, 1971. Grason, J. An Approach to Computerized Space Planning Using Graph Theory. Proceedings of SHARE-ACM-IEEE. Design Automation Workshop, 170179, 1971. Johnson, Timothy, Guy Weinzapfel, John Perkins, Doris C. Ju, Tova Solo, and David Morris. IMAGE: An Interactive Graphics-Based Computer System for Multiconstrained Spatial Synthesis. Cambridge, Mass.: M.I.T., Department of Architecture, 1970 September. Jones, J. Christopher. State of the Art. DMG Newsletter, 5, No. 10, 2, 1971 October.
365;
theNEWMEDIAREADER
Papert, Seymour. Teaching Children to be Mathematicians vs. Teaching About Mathematics, LOGO Memo No. 4. Cambridge, Mass.: Artificial Intelligence Laboratory, M.I.T., 1971c July. Quintrand, Paul. Considrations Gnrales sur Informatique et Architecture. Techniques & Architecture, Srie 33e, No. 4, 6668 (Special), 1971 May. Rittel, Horst. Democratic Decision Making. Summer Session, 71, Architectural Design, 4, 233234, 1972 April. Rorick, Huck. An Evolutionary Architect Wright. Journal of Architectural Education, 26, Nos. 1 and 2, 47, 1971 Winter/Spring. Rubinger, M. State of the Art: A Reply to Christopher Alexander. DMG Newsletter, 5, Nos. 8/9, 4, 1971 AugustSeptember. Rudofsky, Bernard. Architecture without Architects. New York: Museum of Modern Art, 1964. Sanoff, Henry, and Man Sawhney. Residential Livability: A Study of User Attitudes Toward Their Residential Environment. Environmental Design: Research and Practice, William J. Mitchell (editor). Proceedings of the EDRA 3/ar 8 Conference, University of California at Los Angeles, 13.8, 1972 January. Steadman, Philip. Minimal Floor Plan Generation. Cambridge University, Center for Land Use and Built-Form Studies, 1971. Teague, Lavette C., Jr. Network Models of Configurations of Rectangular Parallelepipeds. Emerging Methods in Environmental Design and Planning, Gary T. Moore (editor). Cambridge, Mass.: M.I.T. Press, 1970. Yessios, Christos I., FOSPLAN: A Formal Space Planning Language. Environmental Design: Research and Practice, William J. Mitchell (editor). Proceedings of the EDRA 3/ar Conference, University of California at Los Angeles, 23.9, 1972b January. Wellesley-Miller, Sean. Self-organizing Environments. Design Participation, Nigel Cross (editor). London: Academy Editors, 5862, 1972. Weinzapfel, G. It Might Work, but Will It Help? Proceedings of the Design Activity International Conference, 116, 1973 August. Whitehead, B., and M. Z. Eldars. An Approach to the Optimum Layout of Single-Story Buildings. Architectural Journal (Ba4), 13711380, 1964 June 17.
366
Negroponte, Nicholas, and Leon B. Groisser. URBAN5. Ekistics, 24, No. 142, 289291, 1967a September. Negroponte, Nicholas, and Leon B. Groisser. URBAN5: An On-Line Urban Design Partner. IBM Report, 320-2012. Cambridge, Mass., 1967b June. Negroponte, Nicholas, Leon B. Groisser, and James Taggart. HUNCH: An Experiment in Sketch Recognition. Environmental Design: Research and Practice, William J. Mitchell (editor). Proceedings of the EDRA 3/ar 8 Conference, University of California at Los Angeles, 22.1, 1972 January. Papert, Seymour. Teaching Children Thinking. Teaching Mathematics, No. 58, 1972 Spring. Papert, Seymour. A Computer Laboratory for Elementary Schools. LOGO Memo No. 1. Cambridge, Mass.: Artificial Intelligence Laboratory, M.I.T., 1971a October. Papert, Seymour. Teaching Children Thinking. LOGO Memo No. 2. Cambridge, Mass.: Artificial Intelligence Laboratory, M.I.T., 1971b October.